Protein kinases have become important targets for the design of anti-cancer drugs. Several protein kinase drugs such as Herceptin, Gleevec, and IRESSA have already been approved for treating several forms of cancer. These drugs act on three different protein kinases. Because there are more than 500 protein kinases in human and their various mutant forms are relevant for treating different diseases, there should be more protein kinase-related targets suitable for drug development. The long-term goal of this research is to develop an efficient computational model to help speed up the discovery and development of therapeutic drugs targeting these proteins. The immediate goals of this application aim at further developing a unique five-tier hierarchical computational screening model that balances theoretical rigor and speed at different tiers to shorten the time and expense needed to select the most promising drug candidates from large chemical libraries and to further optimize them for potency, selectivity and drug-like properties. Specific Aim 1 develops a sophisticated continuum solvent model for predicting cLogP and uses it with Lipinski's Rules of Five to evaluate compounds for drug-like properties. Specific Aim 2 improves virtual screening by incorporating protein flexibility in molecular docking and develops an enhanced replica exchange method for refining docking structure for more reliable binding affinity calculation. Specific Aim 3 extends our previous informatics/structural modeling approach to help guide the development of selective protein kinase inhibitors that have fewer side effects. It involves building structural models of the whole family of human protein kinases and on using enhanced conformational sampling techniques (developed in Specific Aim 2) to refine these models well enough for molecular docking and binding affinity calculation. We will also use this whole set of structural models to perform comprehensive comparative analysis to understand better how selectivity can be achieved by protein kinase inhibitors that target the ATP-binding pocket. Specific Aim 4 evaluates the validity of a popular pharmacophore model by performing energy component analysis on all the protein kinase-inhibitor complexes in the Protein Data Bank in order to justify its use in drug design. Specific Aim 5 applies this refined five-tier hierarchical model to identify new anti-cancer drug candidates for the protein kinase targets EGFR, HER2, CDK2, and BCR-ABL. [unreadable] [unreadable] [unreadable]